Fig. 1: Schematic overview linking genetic expression and gradient analyses.

A The median gene expression profiles for 16651 genes were calculated across 200 cortical nodes using Schäefer-200 parcellation. Due to the limited availability of data, we focused on left hemisphere, where gene expression data are available for all six donors. B Microstructural (derived from MT maps) and functional (derived from rs-fMRI) gradients were calculated across the same 200 nodes (for Schäefer 300, 400 parcellations, see Fig. S2) as the gene expression analysis. Color maps represent gradient values. C PLS regression analysis was performed with gene expression profiles as predictors and cortical changes as response variables for n = 10,000 permutations. D PLS assigns weights to each gene indicating its contribution to the overall model for each component. Bootstrapped standard errors were derived and the gene weights were Z-transformed and corrected for multiple comparison using FDR inverse quantile transform correction to account for winners curse. E Genes that were significant after FDR correction (z-score > 1.96) were enriched for tissue types. Genes in PLS component 1 showed significant enrichment; these genes (Table S1 for top ranked 200 genes) are preferentially expressed within occipital and prefrontal brain regions, corresponding to visual and fronto-parietal networks (Yeo7 network). F FC within-network dispersion in the fronto-parietal network (FPN). Red dots, blue dots and the shaded areas represent the dispersion of principal gradients for pre- and post-training sessions, respectively. The shaded area for the post-training session (blue) is larger than that for the pre-training session (red), indicating that the nodes representing functional connectivity similarity within the FPN network are more spread-out after training (i.e. higher network segregation). For 3D illustration see Fig. S1.